Mitochondria are small organelles present within every nucleated mammalian cell. They are the principal source of intracellular energy, in the form of adenosine triphosphate (ATP), which is generated by oxidative phosphorylation (OXPHOS). Mitochondria contain their own DNA- mtDNA - which codes for 13 essential components of the OXPHOS system, along with 22 RNAs required for intra-mitochondrial protein synthesis. Highly deleterious mutations of mtDNA lead to a defect of ATP synthesis and cause multi-system mitochondrial diseases in humans, which typically involve skeletal muscle, the heart, endocrine organs, and the brain. The description of these diseases led to the idea that more subtle variation of mtDNA in the general population might contribute to the aetiology of complex common human disease because of the shared clinical features.

There is emerging evidence that naturally-occurring genetic variation of mtDNA influences the function of mitochondria in humans -possibly by influencing the assembly of OXPHOS protein complexes. It is therefore of great interest that different haplotypes of mtDNA have been associated with type 2 diabetes, hypertension, obesity and the metabolic syndrome, survival following infection, and several neurodegenerative diseases. Many of these studies were small, and the findings have not been replicated. Therefore, the role of common variants of mtDNA in many of these disorders remains unclear.

Commercially-available arrays used for nuclear genome-wide association studies also include a number of mtDNA variants. The precise number varies from platform to platform, but the variants were all chosen based on their frequency in the European population. We have recently developed a method of interpreting this data by imputing the relationship with tightly-associated functional variants through the study of over 1500 complete mitochondrial genomes from population control subjects. We have used this approach to successfully replicate original low-resolution studies associating mtDNA with Parkinson's disease, and by exploiting a large GWAS dataset, we have increased the resolution of the genetic association, deriving the likely mechanism of the association. The data and the tools are therefore in place to study mtDNA genetic variation in the ALSPAC cohort, and to study the relationship with common phenotypes within the cohort.

There are a priori reasons to study the following phenotypes, as examples:

Height

Measures of obesity

Hypertension

Metabolic syndrome

Neurodevelopmental disorders including autism spectrum disorders

Hearing

Vision

It is important to note that, being common genetic variants, in the vast majority of instances, the SNPs will be homoplasmic (ie all the mtDNA in any one individual sample will be identical). In addition, being exclusively maternally inherited, the mtDNA in the child will be identical to the mother. This will allow genotype data from one individual to be reliably used in a phenotypic analysis of another immediate maternal relative.

We would take the following approach:

1. Comparison of allele frequencies in the ALSPAC cohort with other UK datasets including WTCCC3 and the North Cumbria Community Genetics project (data already in our hands). Our prior hypothesis is that the allele frequency will not be significantly different for alleles present at greater than 5% frequency.
2. Principal components analysis, to determine related haplotype groupings, and compare to ethnicity determined by reliable nuclear genetic markers. This would be an essential quality-control (QC) step, because mtDNA is particularly susceptible to the effects of population stratification.
3. Study the relationship between the phenotypes listed above and all variants reaching greater than 95% call rate after appropriate QC. Several statistical approaches could be used. We favour the Tmax function within PLINK because this allows the incorporation of structural relationships within the actual data set, and avoids the need for a harsh Bonferroni-type correction for multiple significances, which is inappropriate for mtDNA alleles. A similar approach will be taken for quantitative trait data.
4. Study the plausibility of interactions existing between nuclear and mitochondrial variation within the ALSPAC sample. We are as yet uninformed as to the existence of such relationships, however the availability of these data will allow for the examination of the possible relationships between these sources of genetic variation and their impact on measured phenotypes.

Based on discussion with the ALSPAC team, we are not requesting any data to be sent out of Bristol. Rather, Dr Gavin Hudson will apply for honorary status with Bristol University, and travel down to Bristol to assist with the analysis, working directly with the Bristol team. In Bristol, Professor Davey Smith, Dr Timpson and PhD student John Kemp will be involved in the analysis of these data and will facilitate this work.